1. Statistical interpretation of ECMWF EPS
Classification of weather patterns with an artificial neural network
Each weather pattern receives a climatological probability for a weather element (rainfall, wind, sunshine,...)

3. Lugano rain > 1 mm/24h
Lugano rain > 10 mm/24h %

5. Probabilities of severe impact weather
Show to the neuronal network the predictors (height, temperature, vorticity, vertical, velocity, stability, wind,...) and the predictand (rainfall, wind,...) by the mean of
supervised learning.

6. Supervised learning, preliminary results
Predictors: H 500 hPa, T850 hPa
Predictand: 24 hour precipitation Lugano
Thresholds: 1, 5, 10, 20, 50 mm
Next figure shows 10mm/24 hours on validation set
Thanks to Ralf Kretschmar

11. Rainfaill Lugano > 50 mm
prob % obs no obs
UNN + PP
PredProbRange obs no obs
[0.00,0.10[
[0.10,0.20[
[0.20,0.30[
[0.30,0.40[
[0.40,0.50[
[0.50,0.60[
[0.60,0.70[
[0.70,0.80[
[0.80,0.90[
[0.90,1.00]
MLP CSC

12. Perspectives
Use predictors more related to rainfall (humidity, vertical velocity,…)
Use regional mean, or better regional maximal rainfall
Translate to ECMWF EPS
Assess strengths and weaknesses of LEPS, EFI, ANN